A new type of artificial intelligence (AI) has been designed to help self-driving cars share useful road data with each other, even if they aren’t connected to the internet. This could make future roads safer and help driverless vehicles make better choices in real time.
The Current Problem with Car Communication
Right now, most autonomous vehicles send their data to central computer servers that analyze and store it. While this system works, it has some issues. It puts a lot of pressure on the servers, and if hackers gain access, a lot of information could be exposed at once. Also, in the current setup, cars need to be close together and allowed to connect before they can swap data.
A Smarter, Safer Way to Share
Researchers have created a new system called Cached Decentralized Federated Learning, or Cached-DFL. Instead of sending data to a central computer, each vehicle stores what it learns while driving inside its own system. When it passes near another car, it can quietly share those lessons without needing any internet or prior setup.
For example, a self-driving car that has faced heavy traffic or spotted a broken traffic light can save that experience. When it meets another car later, it can pass on that knowledge, helping others avoid the same issues—even if they’ve never driven through that area before.
Tested in a Virtual City
To see how well this system works, the team built a computer version of New York City’s Manhattan and added 100 digital cars driving around at random. Each car carried several smart learning models that updated every few minutes. These vehicles didn’t instantly share their data. Instead, they held onto it and waited to connect when another car came within about 100 meters.
This short-range sharing helped cars pick up new insights quickly and make faster, more accurate driving decisions.
Why This Could Change the Future
Experts say this method of learning is helpful because each car only needs to share data with the cars it meets, avoiding the need for large-scale connections. This makes it easier to expand the system without creating overload.
In the future, this technology could also allow cars to connect with other smart things in a city like traffic lights, satellite systems, and even delivery drones. This broader system, often called "vehicle-to-everything," or V2X, could lead to better traffic flow and safer roads.
This approach could also help other smart machines like robots and drones — learn from each other without risking privacy or relying too much on the cloud. It’s a step toward smarter cities and safer automated technology.